Variation of IgG N-linked glycosylation profile in

1 Background: The relationship of IgG glycosylation with diabetes and diabetic 2 nephropathy has been reported, while its role in diabetic retinopathy (DR) remained 3 unclear. We aimed to investigate and validate the association of IgG glycosylation 4 with DR. 5 Methods: We analyzed the IgG N-linked glycosylation profile and identified the 6 specific panel in the discovery population using binary logistics model. Findings were 7 validated in the replication population. The discriminative capacity of IgG 8 glycosylation panel was explored by ROC analysis using cross validation and Brier 9 score. Multiple sensitive analyses were performed on the whole population. 10 Results: 2 IgG glycans (GP15, GP20) and 2 derived traits (IGP32, IGP54) were 11 identified and validated significantly associated with DR (P<0.05), and the adjusted 12 OR were 0.676, 0.671, 1.770, 0.681 in combined population, respectively. The 13 glycosylation panel achieved an average AUC of 0.67 and 0.60 in the discovery and 14 replication population. The association was independent of blood pressure, glucose 15 and lipids, thus improving the ROC and Brier score when the panel added. In addition, 16 the results remained consistent when the controls were re-defined and 1:3 re-matched. 17 Conclusions: IgG glycosylation profile reflecting a pro-inflammatory status were 18 associated with DR. The variation of IgG glycome deserves more attention in the 19 aggravation of diabetes and the underlying mechanism warrants further research.

Type 2 diabetes, characterized by abnormal glycometabolism and impaired insulin 2 function, has become a serious threat to global health. Type 2 diabetes accounts for 3 the vast majority (around 90%) of diabetes worldwide and it is estimated that 171 4 million people have diabetes in 2000 and this number is projected to reach 366 5 million by 2030 [1]. People living with type 2 diabetes are at a higher risk of 6 developing life-threatening complications, such as diabetic retinopathy (DR). About 7 one third of individuals with diabetes have different degrees of DR, which is a 8 common microvascular disease and the leading cause of blindness in the 9 working-aged population [2,3]. In recent years, people with prediabetes, characterized 10 by impaired glucose tolerance and/or impaired fasting glucose, are also increasingly, 11 which signifies a potential risk of the future development of type 2 diabetes and DR 12 [4]. However, the etiological mechanism of the aggravation of diabetes status remains 13 unclear and the potential biological targets related with the onset of DR are urgently 14 needed. 15 Glycometabolism is influenced by the interaction of genetic and environmental 16 factors [5], among which glycosylation is one of the most common and 17 posttranscriptional modifications. The glycans attached to proteins exert crucial 18 biological effect including cellular recognition and molecular pathway regulation [6]. 19 The variation of IgG glycans are mostly investigated and the covalently attached 20 glycans are reported to be associated with the stability of IgG protein and its 21 pro-inflammatory or anti-inflammatory effects [7]. Recently, the variation of IgG 22 glycans are emerging as potential biomarkers and biological pharmacological targets 1 of various metabolic diseases, such as aging [8], dyslipidemia [9], immune disease 2 [10] and type 2 diabetes [11][12]. In fact, type 2 diabetes is accompanied by glucose 3 metabolic disorder and the impaired function of inflammation regulation. Moreover, 4 the IgG glycosylation profiles have been linked with the risk factors of type 2 diabetes, 5 such as obesity [13], blood pressure [14,15] and fasting blood glucose (FBG) [16]. 6 Therefore, it is rational to infer that the specific IgG glycans or traits play an 7 important role in the pathological process of DR. Lemmers et al. [11] and our team 8 [12] have identified the differential IgG glycans between the diabetes population and 9 health controls. However, the biological effect of the IgG glycosylation profile in the 10 development of DR remains unclear. 11 In this study, we aim to investigate the association of the IgG glycosylation with the 12 onset of DR, thus to identify the early glycome biomarkers related with DR. 13

14
Study design and population 15 In 2015, 54 subjects of new-onset DR and 108 matched controls (22 prediabetes and 16 86 diabetes), from the Beijing health management cohort, were enrolled in this study 17 as the discovery population. Subsequently, 54 cases of DR and 108 matched controls 18 (18 prediabetes and 90 diabetes) were recruited in 2016 as the replication population. 19 The Beijing health management cohort is an ongoing population-based study of  IgG glycosylation experiment 10 The glycosylation experiment and analysis composed of four major processes: IgG 11 protein isolation and purification from plasma, N-linked glycans release and 12 fluorescence labeling, glycans quantitative detection, direct glycans and derived traits 13 computation, as described previously [18,19]. In brief, IgG protein was obtained from 14 2ml plasma using 96-well protein G monolithic plates using 1×phosphate buffer 15 saline, 0.1 M formic acid and 1 M ammonium bicarbonate; the N-linked glycans were 16 released from the purified IgG protein at 37℃ using 1.5 units of PNGase F and 17 5×phosphate buffer saline for 18 hours; subsequently, the released glycans were 18 fluorescently labelled using 2-AB at 65℃ for 3 h and isolated with chromatography 19 phase; the direct glycans were quantitatively detected using ultra-performance liquid 20 chromatography platform (Waters, America) and the glycan traits were derived 21 accordingly. 22 Finally, 24 direct glycan peaks (GP) were presented and quantitatively expressed with 1 the percentage of the total integrated peak area. In addition, 54 glycan traits (IGP) 2 were derived to reflect the relative abundance of the specific structure, such as 3 galactosylation, sialylation, bisecting N-acetylglucosamine (GlcNAc), core 4 fucosylation and mannose. The detailed information of each GP and IGP were shown 5 in Appendix Table A.1. The amounts of GP and IGP were normalized by 6 log-transformation and the batch size was considered and corrected for the subsequent 7 analysis. 8 Covariates 9 The demographic characteristics like age and sex were obtained at baseline by 10 questionnaires. The body mass index (BMI) was defined as weight (in 11 kilograms)/height 2 (in meters squared), and was divided into＜25 and ≥ 25. Systolic 12 blood pressure (SBP) and diastolic blood pressure (DBP) were presented as the mean 13 of twice measures on the right arm using sphygmomanometer after resting at least 10 14 min. High blood pressure (HBP) was defined as SBP ≥ 140 or DBP ≥ 90 accordingly. 15 The fasting blood glucose (FBG) was measured after overnight fasting and the 16 postprandial blood glucose (PBG) was measured after 2 hours from the beginning of 17 meals using the glucose oxidase-peroxidase method (Mind Bioengineering Co. Ltd., 18 Shanghai, China). Triglyceride, total cholesterol (TC), high density lipoprotein  Statistical analysis 11 Continuous variables adhering to the normal distribution were represented as the 12 mean ± standard deviation (SD) and the differences between groups were tested by 13 the independent student t tests; otherwise, the interquartile range (P25 -P75) was used 14 and the differences between groups were explored by Mann-Whitney U tests. 15 Categorical variables were presented as n (%), and the differences were tested by the 16 chi-square tests. The box plots were used to show the distribution of IgG glycans and 17 traits between groups. 18 The controls were 1:2 matched based on age, sex and BMI. Binary logistics model 19 was used to identify the IgG glycans and traits associated with the onset of DR in both 20 discovery and replication population. Moreover, the association was explored after 21 confounding covariates (age, sex, BMI, blood pressure, glucose and lipids) adjusted. 22 Further, the discriminative capacity of the differential IgG glycans and traits were 1 explored by ROC analysis using 5-fold cross validation and calibration assessment 2 using Brier score. In addition, the sensitive analyses were performed in the following 3 situations: the ordinal logistics model was used to identify the substantially changed 4 glycans and traits when the controls were re-defined as prediabetes and diabetes; the 5 controls were 1:3 matched and re-analyzed. 6 All reported P values were two-tailed, and P<0.05 was considered statistically 7 significant. All the analyses presented above were performed using the R software 8 (version 3.6.3). 9

10
Demographic and clinical characteristics 11 In the discovery population, the mean age of this population was 61.0 (range from 12 37.0 to 93), involving 131 males (80.9%). In the replication population, the mean age 13 of this population was 60 (range from 27 to 88), involving 135 males (83.3%). The 14 demographic characteristics were similar between the discovery and replication 15 populations. Also, there were no significant difference in HBP, TG, LDLC, HDLC 16 between DR group and the controls, while TC declined in DR group. The detailed 17 information was shown in Table 1. 18

19
In the discovery population, 6 glycans and 9 traits were primarily identified. Then,  Further, discriminative capacity of the glycosylation panel was shown in Table 3  The sensitive analyses were performed on the whole population due to the sample size. 10 On one hand, the controls were separately defined as prediabetes and diabetes. And

21
In this study, we investigated the relationship of IgG glycosylation profile and DR in 22 two matched populations. The panel of GP15, GP20, IGP32, IGP54 was validated to 1 be strongly associated during the pathological process from prediabetes or diabetes to 2 DR, which could also improve the discriminative capacity of the simple model. The results above were largely in consistent with previous studies of IgG 4 glycosylation profiles in type 2 diabetes and its related factors. Previous studies have 5 reported that complex glycan structures with bisecting GlcNAc were highly 6 associated with some inflammatory diseases, reflecting a body status of 7 pro-inflammation [11,27]. IgG proteins were sensitive to biological inflammatory 8 stress and the variation of IgG glycans could reverse its anti-inflammation function 9 [28,29]. Therefore, the substantially increased proportion of the complex glycan 10 structures such as disialylation of fucosylated digalactosylated structures with 11 bisecting GlcNAc may be induced by the biological inflammation in the process of 12 glucose aggravation and DR. In addition, the decreased proportion of galactosylation, 13 accompanied by decreased percentage of sialylation as the sialic acids were attached 14 to the galactose, was thought to strengthen the complement-dependent cytotoxicity 15 (CDC) effect of IgG [30,31]. And the presence of bisecting GlcNAc and lack of core 16 or antennary fucose were thought to strengthen the antibody-dependent cell-mediated 17 cytotoxicity (ADCC) effect of IgG [29,32]. Both the CDC and ADCC effects of IgG 18 were reported to switch its anti-inflammation role to pro-inflammation. Consistently, 19 Lemmers et al. [11] found an glycosylation pattern of decreased galactosylation, 20 sialylation, fucosylation structures and increased bisecting GlcNac structures 21 associated with type 2 diabetes based on a European population. On a further step, we 22 found that the similar IgG glycosylation pattern was associated with the aggravation 1 of diabetes from prediabetes or diabetes to DR in this study. The panel was related 2 with an overall decrease digalactosylated fucosylated structures with and without 3 GlcNAc, with monosialytion or without sialic acid. Moreover, the structures of 4 bisecting GlcNac and disialylation seemed to exert synergetic effect in DR. 5 The strength of our study was that we analyzed the variation of IgG glycosylation 6 profiles and identified the IgG glycans and traits associated with DR for the first time. Ethics approval and consent to participate 13 The study followed the guidelines of the Helsinki Declaration, and was approved by 14 the Ethics Committees of Capital Medical University. 15 Consent for publication 16 All participants have given the consent for publication.

17
Availability of data and materials 18 The datasets used and/or analysed during the current study are available from the 19 corresponding author on reasonable request. 20 Competing interests 21 The authors declare that they have no competing interests.  applicable   15   16   17  18  19  20  21  22  23  24  25  26  27  28  Tables   1  2   Table1   3 The characteristics of participants in the discovery and replication populations.